- Researchers in Japan have unveiled a robot that has been trained to carefully peel bananas without crushing them.
- The dual-armed robot has managed to peel the banana successfully 57% of the time, providing hope and insight into a future where machines can perform more nimble operations.
- University of Tokyo researchers Heecheol Kim, Yoshiyuki Ohmura and Yasuo Kuniyoshi performed a “deep imitation learning” method on the robot where the researchers would repeat the action until the robot was able to successfully learn and imitate it.
- After 13 hours of deep imitation learning, the robot was able to successfully replicate the action of peeling a banana.
- Given labor shortages in Japan, the researchers are hoping that the robot can alleviate the problem by working at processing factories.
Researchers at the University of Tokyo in Japan have trained a dual-armed robot to successfully peel a banana without crushing it, giving them hope of developing better-trained machines to perform simple “human” tasks.
A video by the researchers shows the robot carefully picking up a banana and peeling its outer layer utilizing both mechanical arms.
Researchers Heecheol Kim, Yoshiyuki Ohmura and Yasuo Kuniyoshi trained the robot using a method called “deep imitation learning,” where they would demonstrate the action of peeling a banana hundreds of times until the robot was able to successfully memorize and imitate the action.
The training process took the researchers about 13 hours before the robot was able to reach a consistent success rate.
Now managing to successfully perform the task 57% of the time, the robot provides a hopeful outlook on a future in which better-trained robots can perform more nimble, basic human tasks.
Kuniyoshi also believes that the ongoing labor shortage in Japan may be resolved with the help of these robots, as he foresees them performing human labor tasks at processing factories.